Valve on the Importance of Using Sweat to Make Games Better

By Dennis Scimeca

Mike Ambinder, an experimental psychologist from Valve Software, presented an utterly fascinating lecture at GDC yesterday on how the studio uses biofeedback to enhance gaming experiences.

Valve is renowned for its trailblazing achievements in storytelling, game mechanics, and digital distribution, but what is less popularized outside of accessible game design circles is their pioneering work in new technologies. Like a real-life Aperture Laboratories without the torture, Valve is always tinkering with new techniques to improve our gaming experiences. Want to find out how they tinker and sweat? Read on to find out.

“Current control schemes only provide one dimension of input,” said Ambinder. Designers map player intent to onscreen action, and ignore other aspects of cognition (thought), like player sentiment. By measuring or adding biofeedback data which can measure that sentiment, Valve is attempting to create more immersive games. This was an extremely dense panel with far too much information to adequately cram into a single blog post, and I heartily recommend you view the video of this panel when it’s uploaded to the GDC website. But one of the three biofeedback experiments that Ambinder presented more than adequately demonstrated the potential for these techniques.

Valve wanted to measure skin conductance level, or SCL, of subjects playing Left 4 Dead 2. By placing two metal contacts close to each other on a player’s hand and creating a viable current, Valve could measure the electrical resistance of the player’s skin. The more the player sweats, the more current gets through. SCL is an excellent biofeedback tool because it provides very quick response to stimuli, and by way of demonstrating the point, Ambinder showed a video measuring SCL responses of a subject playing L4D2.

A line chart in the upper-right-hand corner of the screen showed the data, and turned red whenever the SCL spiked. Almost every time the player heard a threatening noise, passed by an open door, or got into a hairy fight with the infected, the chart turned beet red and the readings shot way up. The almost 1:1 response of threat to reaction was compelling evidence of how well this technique works.

By analyzing this data, Valve could produce gameplay events that produced greater user arousal, or excitement, and the developers could gain rudimentary insight into which events elicit more enjoyment than others. If this data could be hooked directly into the AI Director in Left 4 Dead, which modifies spawns, health, weapon placements, and boss appearances, the gaming experience could tailor itself to the player’s reactions in excitingly effective ways.

Like I said, this was just the barest fraction of what Ambinder managed to condense in a panel that was only an hour long. Perhaps next year GDC can be convinced to give him two or three hours, because this was easily, for me, one of the most fascinating panels of the conference.